期刊
IEEE EMBEDDED SYSTEMS LETTERS
卷 14, 期 3, 页码 155-158出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/LES.2021.3137335
关键词
Approximate computing; convolution neural network; multiply and accumulate (MAC)
类别
资金
- National Natural Science Foundation of China (NSFC) [61834006, 62025404, 62122076, 62104229]
The article discusses the significance of multiply and accumulate (MAC) in convolutional neural network accelerators and proposes an approximate MAC unit design that takes into account the statistical features of input data to achieve a balance between latency, power, and accuracy.
Multiply and accumulate (MAC) composed of a set of multipliers and one reduction dominates the latency and power of convolutional neural network (CNN) accelerators. Existing approximate multipliers reduce latency and power at a tolerable drop in accuracy, without considering the data distribution (implicitly assuming that data are uniformly distributed). This letter discloses that practical CNNs' activations and weights are usually Gaussian-like distributed, and the bits of quantized activations and weights are typically not with a probability of 0.5. Thus, we propose an approximate MAC unit design by taking into account the statistical features of input data, to achieve a balanced tradeoff among latency, power, and accuracy. The extensive experiments show that our proposed MAC unit design provides much higher accuracy than state-of-the-art approximate circuits, while the latency, area, and power are similar.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据